SALAMANDRA 54(4) 259–268 30 OctobSeyr n2o0n1y8my oIfS tSwNo 0B0r3o6o–k3e3si7a5 taxa Polymorphism and synonymy of Brookesia antakarana and B. ambreensis, leaf chameleons from Montagne d’Ambre in north Madagascar Mark D. Scherz1,2, Frank Glaw1, Andolalao Rakotoarison3, Melina Wagler2 & Miguel Vences2 1) Zoologische Staatssammlung München (ZSM-SNSB), Münchhausenstr. 21, 81247 München, Germany 2) Braunschweig University of Technology, Division of Evolutionary Biology, Zoological Institute, Mendelssohnstr. 4, 38106 Braunschweig, Germany 3) Zoologie et Biodiversité Animale, Université d‘Antananarivo, BP 906, Antananarivo, 101 Madagascar Corresponding author: Mark Scherz, e-mail: [email protected] Manuscript received: 9 May 2018 Accepted: 3 August 2018 by Jörn Köhler Abstract. We examine the taxonomic status of two Malagasy leaf chameleon taxa, Brookesia antakarana Raxworthy & Nussbaum, 1995 and B. ambreensis Raxworthy & Nussbaum, 1995, integrating morphological and genetic evidence. Specimens assigned to these species occur in syntopy in Montagne d’Ambre, northern Madagascar, and were originally described based on differences in the shape of their pelvic shields. We found that the shape of these shields falls on a con- tinuous spectrum, and detected only weak differences between the two taxa in a few other morphological features, all of which were correlated with shield length. Members of the two taxa (as assigned based on pelvic shield morphology) also showed extensive haplotype sharing in one nuclear and one mitochondrial marker. We conclude that at present there is no convincing evidence that these species are distinct, and act as first revisers in the sense of the International Code of Zoo- logical Nomenclature to place B. ambreensis into the synonymy of B. antakarana. Key words. Squamata, Chamaeleonidae, morphology, molecular genetics, taxonomy, Amber Mountain, Antsiranana. Introduction Nussbaum (1995), but the values given in the diagnoses of the species (12 in B. ambreensis vs. 12–13 in B. antakara In their work on the diversification patterns of Mada- na) do not agree with those given in their key (11 in B. am gascar’s dwarf chameleons (genus Brookesia Gray, 1865), breensis vs. 12–13 in B. antakarana), leaving the value of Townsend et al. (2009) showed that specimens from Mon- this character unclear. A further character, the number of tagne d’Ambre in northern Madagascar assigned to the two dorsolateral spines on the tail, was mentioned in the de- locally endemic species Brookesia antakarana Raxworthy scriptions of the holotypes of both species and the diagno- & Nussbaum, 1995 and B. ambreensis Raxworthy & Nuss- sis of B. ambreensis, but was not included in the diagnosis baum, 1995 were highly similar genetically. Their recon- of B. antakarana or in the key; its potential relevance was struction of the phylogeny of these chameleons found the highlighted however in Glaw & Vences (2007). Finally, two taxa to be interdigitated, and characterised by compar- colouration was apparently different between the two taxa, atively low genetic divergences. In light of this finding, it is with B. ambreensis described by Raxworthy & Nussbaum pertinent to review the taxonomy of these species, and the (1995) as dorsally ‘unmarked, or with a thin dark brown question of whether or not they are synonymous. vertebral line,’ while B. antakarana was described as dor- In their original descriptions, Raxworthy & Nuss- sally having ‘three, dark brown, blunt chevrons’. baum (1995) considered these two species to be strongly Given the syntopic occurrence of these two species (al- differentiated, predominantly on the basis of the ‘pelvic titudinal ranges of 650–1200 m a.s.l. in B. antakarana and shield’, which is ‘diamond shaped’ in B. antakarana versus 650–1150 m a.s.l. in B. ambreensis according to Raxwor- ‘no well-defined pelvic shield’ in B. ambreensis. Addition- thy & Nussbaum 1995, and found in syntopy by us), and ally, the number of dorsolateral pointed tubercles (here the evidence of Townsend et al. (2009) that they are ge- termed ‘dorsolateral spines’) would be a candidate distin- netically non-assortative, a revision of their taxonomic sta- guishing character according to the key of Raxworthy & tus is necessary. © 2018 Deutsche Gesellschaft für Herpetologie und Terrarienkunde e.V. (DGHT), Mannheim, Germany Available at http://www.salamandra-journal.com 259 Mark D. Scherz et al. Materials and methods DNA sequencing and analysis of sequences Sampling and terminology DNA was extracted following standard salt extraction pro- Specimens were sought at night in Montagne d’Ambre tocols, using proteinase K digestion in a concentration of along trails and within the forest on thin branches and 10 mg/ml (Bruford et al. 1992). For analyses of mitochon- leaves by torchlight. Field identification of individuals was drial divergence, we targeted fragments of one mitochon- based primarily on the condition of the pelvic shield. Some drial gene: NADH dehydrogenase subunit 2 (ND2). To un- individuals were collected as specimens, and euthanized derstand concordance of mitochondrial and nuclear DNA with concentrated MS 222, fixed in 90% ethanol and stored differentiation, we focused on one nuclear gene segment in 70% ethanol. Tissue samples (muscle) were placed in of the recombination-activating gene 1 (Rag-1). For prim- 96% ethanol for genetic study. Specimens were deposited ers and protocols, see Townsend et al. (2009) and Tolley at the collection of the Université d’Antananarivo, Mention et al. (2013) for ND2, and Rakotoarison et al. (2015) for Zoologie et Biodiversité Animale, Madagascar (UADBA) Rag-1 (nested approach). PCR products were sequenced and in the Zoologische Staatssammlung München, Ger- directly using an automated DNA sequencer (ABI 3130 many (ZSM). Reference is made to specimens in the Uni- XL, Applied Biosystems). Quality control of sequences was versity of Michigan, Museum of Zoology (UMMZ). Field carried out using CodonCode Aligner (Codon Code Cor- numbers refer to Ronald A. Nussbaum (RAN), Frank poration). For sequence alignment, as well as calculation Glaw and Miguel Vences (FG/MV, FGZC). of uncorrected p-distances between sequences, we used MEGA7 (Kumar et al. 2016). Newly determined sequences were deposited in GenBank with the following accession Morphology and morphometrics numbers: MH683055–MH683089. The two markers were analysed separately because our Callipers were used to measure the following variables primary objective was to obtain evidence from unlinked to the nearest 0.1 mm: Snout–vent length (SVL), body loci (mitochondrial versus nuclear) for genetic differentia- height (BH), tail length (TAL), eye diameter (ED), orbit- tion of lineages, which would provide support for their sta- al crest to eye distance (OC), mouth length (ML), upper tus as distinct species. Haplotypes of the Rag-1 fragment arm length (UAL), lower arm length (LAL), thigh length were inferred using the PHASE algorithm (Stephens et al. (TL), shank length (SL), pelvic shield width (SHW), pel- 2001) implemented in DnaSP software, Version 5.10.3 (Li- vic shield length (SHL), and pelvic shield diagonal (SHD). bra do & Rozas 2009). From the phased Rag-1 sequenc- Relative measures used in statistics discussed below were es and the unphased ND2 sequences, we reconstructed a calculated by dividing these measures by SVL. Further- Maximum Likelihood tree with Jukes-Cantor substitution more, the following scale counts were done: scales around model (the most simple available model, chosen to avoid midbody (SAR), number of supraocular crest spines over-parametrisation considering the very few mutations (SCS), scales along posterior casque limit (CLS), infrala- in the data set) in MEGA7 (Kumar et al. 2016) and en- bial scales (ILS), scales from casque to eye (SCE), number tered this tree together with the alignment in the software of dorsolateral spines including the sacral spine following Haploviewer, written by G. B. Ewing (http://www.cibiv. Raxworthy & Nussbaum (1995) (DLS), and number of at/~greg/haploviewer), which implements the methodo- hand scales between finger tips (NHS). Finally, we cod- logical approach of Salzburger et al. (2011). ed presence (+) or absence (–) of tail spines (TLS); these were not counted because they decrease in size toward un- detectability posteriorly, and are generally either present Results over the entire length of the tail or wholly absent. For a Morphological analyses schematic diagram of investigated characters, see Figure 1. Counts and measurements were carried out by student Voucher specimens available from our fieldwork con- assistants (and some repeated and verified by MDS). Each sisted of eight individuals morphologically identified as count or measurement was carried out by a single person Brookesia ambreensis (four males and four females) and over 1–2 days, without a priori knowledge of the identity 11 individuals identified as B. antakarana (nine males and of the specimens, to avoid inter-individual differences in two females), all preserved in the ZSM (Table 1). Three ad- count interpretations. Morphometric and meristic data ditional individuals from the UADBA collection, identi- were analysed in R 3.4.3 (R Core Team 2014). Homoge- fied in the field as either B. ambreensis (two specimens) neity of variance of variables was assessed with Levene’s or B. antakarana (one), were included in our molecular tests. Residuals of Generalised Linear Models (GLMs) analysis but were not available for morphological analysis. were assessed for normality with Q-Q plots and Shapiro- Overall, ND2 sequences were obtained for six B. ambreen Wilk tests. GLM denominator degrees of freedom were sis and eight B. antakarana, whereas Rag-1 sequences were calculated based on Satterthwaite’s approximation. Prin- obtained for seven B. ambreensis and eight B. antakarana cipal Component Analyses were conducted with scaled (Table 1). In addition, morphometric measurements were and centred log-transformed data to reduce magnitude also taken from sympatric B. stumpffi for comparative pur- and skew biases. poses (see revised diagnosis, below). 260 Synonymy of two Brookesia taxa Morphological analysis of the main character diagnos- holotypes of the two species fell into these two categories, ing B. ambreensis and B. antakarana, the pelvic shield, re- in agreement with the original description (Raxworthy & vealed strong differences in the individuals assigned to Nussbaum 1995). However, several individuals had inter- these two species (Fig. 2), because the morphology of the mediate character states, e.g. ZSM 1504/2008 (Fig. 2). pelvic shield was the primary feature used to assign speci- We quantified the length and width of the pelvic shield mens to either species. Specimens assigned to B. antaka by three morphometric measurements (SHL, SHW, rana have a clearly defined rhomboid of roughly equiva- SHD). As expected given their trigonometric relation- lent length and width, whereas in B. ambreensis, the ‘shield ship (see Fig. 1), SHL and SHD were very strongly corre- area’ was more elongated posteriorly, extending onto the lated (Pearson’s product-moment correlation, correlation tail, and thus often not discernible as a clear shield. The = 0.98, p < 0.0001), but both were also moderately cor- Figure 1. Morphometric measurements and scale counts taken for specimens of Brookesia. Abbreviations: BH – body height, ILS – infralabial scales, LAL – lower arm length, ML – mouth length, OC – orbital crest to eye distance, SCE – scales from casque to eye, SCS – number of supraocular crest spines, SHL – pelvic shield length, SHD – pelvic shield diagonal, SHW – pelvic shield width, NHS – number of hand scales between finger tips, ED – eye diameter, SL – shank length, SVL – snout–vent length, TAL – tail length, TL – thigh length, TOL – total length, UAL – upper arm length. 261 Mark D. Scherz et al. Table 1. Morphological measurements (in mm) of Brookesia species from Montagne d’Ambre National Park (including former Forêt d’Ambre Special Reserve). See Materials and Methods for abbreviations. GenBank accession numbers are provided for ND2 and Rag-1 (marker not sequenced if accession number missing). Catalogue Field number Morphological ND2 Rag-1 SVL TAL ED OC ML UAL LAL TL SL SHW SHL SHD number species identification Males ZSM 1029/2003 B. ambreensis – – 47.0 35.0 4.8 1.7 9.5 10.8 13.3 8.6 11.7 7.2 5.6 6.5 ZSM 1030/2003 FG/MV 2002.3086 B. ambreensis MH683056 MH683088 49.7 35.0 4.5 1.7 9.7 8.2 12.2 8.7 11.1 7.1 7.8 8.6 ZSM 2090/2007 FGZC 1071 B. ambreensis MH683061 MH683077 46.2 30.2 4.3 1.8 8.6 7.5 11.2 8.7 10.5 7.3 5.4 6.8 ZSM 226/2004 FGZC 445 B. ambreensis FJ975197 MH683078 43.1 27.6 4.2 0.5 7.9 8.0 10.2 7.4 9.6 5.8 6.4 7.0 ZSM 1031/2003 FG/MV 2002.2374 B. antakarana – MH683079 49.2 32.4 4.5 1.8 9.7 7.8 11.4 9.5 10.4 6.9 2.4 4.1 ZSM 1032/2003 FG/MV 2002.2375 B. antakarana MH683065 MH683080 47.4 29.1 5.6 1.3 9.1 8.9 11.5 8.3 10.2 6.5 3.2 4.2 ZSM 1003/2003 FG/MV 2002.3018 B. antakarana MH683055 MH683081 43.1 31.0 4.6 1.6 9.2 8.2 10.8 8.2 10.2 7.1 3.9 4.9 ZSM 1034/2003 FG/MV 2002.3087 B. antakarana MH683058 MH683086 37.6 28.1 3.9 1.3 8.2 7.9 11.5 8.1 9.4 6.4 3.9 5.2 ZSM 2092/2007 FGZC 1074 B. antakarana MH683063 MH683082 45.5 28.3 4.1 1.5 8.8 7.3 11.3 8.2 9.5 6.1 3.9 4.7 ZSM 2093/2007 FGZC 1075 B. antakarana MH683064 MH683083 45.7 26.2 4.5 1.0 8.7 6.2 10.8 7.6 10.0 6.6 3.4 4.5 ZSM 225/2004 FGZC 444 B. antakarana FJ975195 MH683084 45.1 29.3 4.9 0.9 8.9 8.9 11.0 7.7 10.8 6.8 3.8 4.7 ZSM 234/2004 FGZC 456 B. antakarana FJ975196 MH683085 54.3 33.3 5.3 0.9 9.4 8.8 11.8 7.5 10.7 6.9 3.8 4.9 ZSM 1661/2012 FGZC 4913 B. antakarana – – 48.6 29.7 4.5 0.9 8.0 7.6 10.6 7.3 8.7 6.6 3.0 4.5 ZSM 2095/2007 FGZC 1079 B. stumpffi MH683071 – 49.8 33.4 5.7 1.5 9.7 8.7 11.3 9.6 12.4 6.4 4.7 5.8 ZSM 1681/2012 FGZC 4920 B. stumpffi – – 41.5 29.7 3.8 1.4 7.6 7.2 9.8 8.7 10.0 6.0 4.4 5.2 Females ZSM 1028/2003 FG/MV 2002.2378 B. ambreensis MH683057 MH683075 46.7 40.2 4.4 2.1 10.8 10.4 12.5 11.3 12.3 10.6 9.4 10.8 ZSM 2088/2007 FGZC 1068 B. ambreensis MH683059 MH683089 51.2 33.0 4.5 1.4 10.5 9.3 12.0 9.5 11.5 8.0 5.9 7.3 ZSM 2089/2007 FGZC 1070 B. ambreensis MH683060 MH683076 51.8 37.0 4.6 1.7 11.0 10.2 13.0 10.6 13.3 8.2 6.7 8.0 ZSM 1504/2008 FGZC 1858 B. ambreensis – MH683087 45.9 31.0 5.2 1.6 10.2 9.9 13.9 10.9 12.2 8.3 6.9 7.5 ZSM 1033/2003 FG/MV 2002.2380 B. antakarana – – 60.8 39.6 5.1 1.5 11.1 10.4 14.1 10.0 12.8 8.5 3.5 5.5 ZSM 2091/2007 FGZC 1073 B. antakarana MH683062 – 48.1 37.0 4.6 1.4 10.6 9.2 12.8 8.2 11.0 8.6 3.2 5.2 ZSM 2094/2007 FGZC 1077 B. stumpffi MH683070 – 51.8 38.1 4.1 1.9 9.9 10.7 12.9 11.1 12.4 7.3 6.3 7.3 ZSM 2165/2007 FGZC 1234 B. stumpffi – – 47.3 33.9 4.2 1.6 9.4 9.0 13.1 10.4 12.2 6.9 4.6 5.7 ZSM 2166/2007 FGZC 1235 B. stumpffi – – 54.0 37.7 4.2 2.0 11.3 9.9 13.0 10.2 12.3 7.0 5.0 6.1 Catalogue number Morphological species identification SAR SCS CLS ILS SCE DLS TLS NHS Males ZSM 1029/2003 B. ambreensis 165 13 29 19 11.5 13 – 31.0 ZSM 1030/2003 B. ambreensis 200 16 24 22 15.0 12 – 31.5 ZSM 2090/2007 B. ambreensis 181 11 23 19 11.0 12 – 25.5 ZSM 226/2004 B. ambreensis 210 16 24 20 11.0 12 – 26.0 ZSM 1031/2003 B. antakarana 221 14 22 20 9.0 13 + 26.0 ZSM 1032/2003 B. antakarana 189 13 21 18 10.0 12 – 23.0 ZSM 1003/2003 B. antakarana 184 13 25 19 11.0 12 + 26.0 ZSM 1034/2003 B. antakarana 199 12 21 19 10.5 12 – 25.0 ZSM 2092/2007 B. antakarana 194 14 22 20 9.0 12 + 28.0 ZSM 2093/2007 B. antakarana 205 10 22 18 9.0 12 + 24.5 ZSM 225/2004 B. antakarana 242 13 26 21.5 9.0 12 + 28.5 ZSM 234/2004 B. antakarana 206 14 22 20 12.0 12 + 21.0 ZSM 1661/2012 B. antakarana 214 14 19 19 9.0 11 – 30.0 ZSM 2095/2007 B. stumpffi 193 15 22 19 13.0 10 + 25.0 ZSM 1681/2012 B. stumpffi 196 14 20 18.5 11.5 9 + 21.5 Females ZSM 1028/2003 B. ambreensis 227 12 16 20 9.0 12 – 25.5 ZSM 2088/2007 B. ambreensis 226 15 29 19.5 12.0 13 + 31.5 ZSM 2089/2007 B. ambreensis 235 14 18 20 12.0 12 + 33.5 ZSM 1504/2008 B. ambreensis 228 12 27 23 11.0 12 – 24.5 ZSM 1033/2003 B. antakarana 204 13 23 21 11.0 12 – 25.0 ZSM 2091/2007 B. antakarana 204 12 24 19 8.0 12 + 26.5 ZSM 2094/2007 B. stumpffi 227 15 23 20.5 14.5 9 + 24.0 ZSM 2165/2007 B. stumpffi 203 14 27 22 10.0 10 + 21.0 ZSM 2166/2007 B. stumpffi 198 14 21 22 13.0 10 + 23.0 262 Synonymy of two Brookesia taxa related with SHW (Pearson’s product-moment correla- Relative shank length (SL) was also found to be signifi- tion, SHW~SHD = 0.59, p = 0.0075; SHW~SHL = 0.51, cantly different (F = 5.8017, p = 0.02768), and relative 1,16.95 p = 0.0273), suggesting that the length of the pelvic shield thigh length (TL) and upper arm length (UAL) were close (of which both SHL and SHD are metrics) is related to its to statistical significance (F = 3.7672, p = 0.06904 and 1,16.70 width (SHW). Statistical analysis comparing the differenc- F = 3.2903, p = 0.08803, respectively) between speci- 1,16.40 es in characters between specimens assigned to B. anta mens assigned to either species in GLM analysis; in each karana and B. ambreensis with sex as a random factor case, specimens assigned to B. ambreensis tended to have (generalised linear model [GLM]: Character ~ Species + higher values than B. antakarana. Each of these measures (1|Sex)) yielded highly significant results in SHL and SHD correlated significantly with SHL and thereby also SHD (SHL: F = 51.674, p = 1.515e-06; SHD: F = 47.74, (Pearson’s product-moment correlations: SL~SHL, corre- 1,17.00 1,17.00 p = 2.523e-06) but not SHW (F = 0.6356, p = 0.4366). lation = 0.65, p = 0.0026; UAL~SHL, correlation = 0.50, 1,16.61 The same results were found under non-parametric Wil- p = 0.028; TL~SHL, correlation = 0.62, p = 0.0046). None coxon U-tests between males of the two species (insuffi- of these differences were highlighted as significant in Wil- cient female specimens to perform direct comparisons), coxon U-tests between males of the two species after Bon- which remained significant after Bonferroni correction ferroni correction. No relationship was found between as- for multiple comparisons (both SHL and SHD: W = 36, signed species and the number of dorsolateral spines (chi- p = 0.0364). Again, this finding is unsurprising given our a square test, χ² = 1.497, df = 2, p = 0.4731); both groups had priori assignment of specimens to either species based on individuals with 12 and 13 dorsolateral spines, though one the shape of the pelvic shield. specimen assigned to B. antakarana also had 11 dorsolater- Figure 2. Pelvic shield region of specimens assigned to B. ambreensis and B. antakarana, separately for males and females. Bold and underlined voucher numbers (from the UMMZ collection) refer to the holotypes of the respective species. 263 Mark D. Scherz et al. al spines. Furthermore, specimens assigned to either spe- Table 2. Principal component loadings, standard deviation, and cies had dorsolateral spines on their tails, although these proportion of variance, for principal component analysis includ- were present in more specimens of B. antakarana (6/10) ing shield-related values. than B. ambreensis (2/8). Principal component analysis (PCA) of scaled and cen- PC1 PC2 PC3 tred log-transformed measurement and count data (SVL, Standard Deviation 2.6010 1.8075 1.5515 TAL, ED, OC, ML, UAL, LAL, TL, SL, SHW, SHL, SHD, Proportion of Variance 0.3560 0.1719 0.1267 SAR, SCS, CLS, ILS, SCE, DLS, NHS) showed limited dis- Cumulative proportion 0.3560 0.5280 0.6546 placement between specimens assigned to either taxon on SVL 0.15297 0.13788 -0.43115 the first through third principal components (PCs; Fig. 3). Together these three PCs accounted for 66% of the varia- TAL 0.34646 0.01036 -0.16909 tion in the data (Table 2); eight PCs were required to ac- ED 0.12717 0.24622 -0.37442 count for over 95%. The first PC was size related, most OC 0.22297 0.26572 0.26007 strongly loaded by tail length, but the second and third ML 0.30086 0.24223 -0.13377 were loaded most strongly by pelvic shield measurements, UAL 0.25849 -0.02008 -0.12847 scale counts, ED, OC, and SVL (Table 2). What clustering LAL 0.32142 0.11852 0.07269 there was (no overlap between convex hulls in PC1 vs. PC2, TL 0.22741 0.20187 0.19722 Fig. 3) was thus apparently related largely to body size and SL 0.33953 0.08293 -0.03436 especially tail length, as well as pelvic shield shape. Any SHW 0.26571 0.19746 0.00134 semblance of assortment disappeared completely when SHL 0.19576 -0.40691 0.18435 pelvic shield characters were excluded from the analysis SHD 0.21154 -0.39300 0.18255 (Fig. 3). SAR -0.17557 -0.09299 -0.41640 SCS 0.07802 -0.31595 -0.34363 Chromatic variation CLS 0.24353 -0.14910 0.11555 ILS 0.14853 -0.31884 -0.28586 Examination of photos in life of specimens involved in SCE 0.27027 -0.23952 0.01547 this study (Fig. 4) shows that two colour morphs are rec- DLS 0.03426 0.13170 0.18110 ognisable, namely one in which clear, almost right-angled NHS 0.12987 -0.25061 0.10509 chevrons are present on the dorsum and the overall col- ouration is mottled and greyish (Figs 4a, c, e), and a sec- ond in which the vertebral stripe is light in colour with ei- observed in Rag-1, with a maximum number of 11 steps be- ther longer, acute chevrons, or lacking chevrons entirely, tween haplotypes in the network (Fig. 5) and sharing in the and the overall body colouration is more uniformly brown two most common out of nine haplotypes. (Figs 4b, d, f–h). The right-angled chevron morph matches A reanalysis of the DNA sequence data provided by the description of B. antakarana, while the more uniform Raxworthy et al. (2002) agreed with this intriguing pic- colour morph with a light vertebral stripe is similar to that ture. In the mitochondrial ND4 gene the distances between described for B. ambreensis. However, these colour morphs samples of B. antakarana and B. ambreensis amounted to are unrelated to pelvic shield morphology; Fig. 4f shows only 0.8–1.7% uncorrected p-distance, corresponding to a specimen with a well-defined diamond-shaped pelvic 7–16 substitutions in the 922 bp fragment sequenced by shield with highly acute chevrons on a light dorsal stripe; Raxworthy et al. (2002). In contrast, this gene typically that is, the pelvic shield morphology of B. antakarana with shows quite high inter-specific divergences among many the colouration of B. ambreensis. squamate species, for instance 9–19% uncorrected pairwise distances among various Brookesia species on the basis of sequences from Raxworthy et al. (2002). One sequence Genetic results of a B. ambreensis paratype (AF443238; field number RAN 38125, corresponding to UMMZ 200074) differed by only The analysed ND2 fragment, after trimming of stretches 7 substitutions from the single sequence of a B. antakara at the beginning and end with missing data, had a length na paratype (AF443249; RAN 38057 = UMMZ 203636) but of 411 base pairs. All sequences were very similar, with a by 11 substitutions from a second B. ambreensis paratype maximum number of eight nucleotide substitutions. Un- (AF443240; RAN 38476 = UMMZ 200077). corrected pairwise distances were 0.0–1.7% within B. am breensis, 0.0–1.7% within B. antakarana, and 0.0–1.7% be- tween B. ambreensis and B. antakarana. The haplotype net- Discussion work (Fig. 5) reflects this situation and groups all ND2 se- quences into a single network, with a maximum of eight It is clear from our results that there are no concordant steps, and with four of six haplotypes shared between morphological, chromatic, or genetic differences among B. ambreensis and B. antakarana. A similar situation was the two taxa Brookesia ambreensis and B. antakarana; al- 264 Synonymy of two Brookesia taxa though there was a significant difference in the length and clude that this is not a useful character for distinguishing width of the pelvic shield, these differences are a result of a these taxa. However, we note that there is a clear segrega- priori use of this character to assign specimens to each tax- tion between these two taxa and the partially sympatric on, and do not therefore constitute independent evidence B. stumpffi, which has 9–10 dorsolateral spines (though we of morphologically distinct groups of individuals. Other note that our sample size of B. stumpffi is small at n = 5; Ta- weak differences between these species correlate with the ble 1). Tail spines also appear to be polymorphic in speci- length of the pelvic shield, suggesting that some character- mens assigned to either taxon, and are not assortative. istics may be linked to variation in this structure, possibly The lack of genetic differentiation in nuclear and mito- through pleio tropic genetic effects. The taxa did not clus- chondrial genes, combined with (i) weak morphological ter strongly in principal component analyses, and lost any differentiation that is strongly related to a priori methods clustering when pelvic shield characters were omitted. The used in species assignment, which (ii) occurs on a con- differences in pelvic shield length cannot be explained by tinuous gradient and not two absolute classes as original- sexual dimorphism of a single species, as they are unrelat- ly proposed, suggests that the specimens we have exam- ed to sex of the specimens (see Fig. 2). ined belong to a single species with variable pelvic shield As mentioned above, the number of dorsolateral spines morphology. Further support for the lack of distinction has been suggested to be useful for distinguishing B. an between the two taxa comes from the three other poten- takarana and B. ambreensis, but the values given in the di- tially diagnostic characters mentioned in the introduction: agnosis (and holotype description) of B. ambreensis do not (1) the number of dorsolateral spines does not differ be- match those given in the key of Raxworthy & Nussbaum tween them, (2) the presence and absence of spines on the (1995). We found no relationship between the pelvic shield tail is not diagnostic, and (3) colouration is independent of morphology and number of dorsolateral spines, and con- pelvic shield morphology. Figure 3. Principal component analysis of the morphology of male specimens assigned to Brookesia antakarana and B. ambreensis. Parenthetical values indicate the percentage of the variation explained by the principal component. Weak segregation of B. antakarana and B. ambreensis is evident in PCs 1–3 when total morphology is considered (above), but vanishes when pelvic shield characters are omitted (below). 265 Mark D. Scherz et al. In summary, it is apparent that these two names repre- Nomenclature Article 24.2.1 with regard to their priority sent only a single species, and therefore require synonymi- (ICZN 1999). We here act as ‘first revisers’ in the sense of sation. As the two names were erected together in the same the Code Article 24.2.1, and designate B. ambreensis as jun- paper, they are seen as having been published ‘simultane- ior synonym of B. antakarana. We take this decision on the ously’ in the sense of the International Code of Zoological basis that several species endemic to Montagne d’Ambre Figure 4. Specimens of Brookesia antakarana/ambreensis photographed in Montagne d’Ambre between 1994 and 2012. (A) Probably ZSM 2091/2007 (FGZC 1073); (B) unidentified individual photographed in 2007; (C) unidentified individual photographed in 2003; (D) probably ZSM 2030/2003 (FGMV 2002-3086); (E–F) uncollected individuals photographed in 1994; (G–H) unidentified individual photographed in 2012. 266 Synonymy of two Brookesia taxa and the surrounding area carry the specific epithet ‘am eye is less than the eye diameter; colouration highly vari- breensis’ or a variation thereof, e.g. the chameleons Calum able, it can include a light dorsolateral stripe or dark chev- ma ambreense (Ramanantsoa, 1974) and C. amber Rax- rons; SVL up to 37.6–54.3 mm in males, 45.9–60.8 mm in worthy & Nussbaum, 2006, and the frog Mantidactylus females. ambreensis Mocquard, 1895. On the contrary, no other Brookesia antakarana can be distinguished from the reptiles or amphibians yet have the name ‘antakarana’, and generally similar species B. brygooi, B. decaryi, B. bonsi, in consequence the risk of confusion is considerably lower B. stumpffi, B. griveaudi, B. valerieae, and B. lineata as fol- with this name than the alternative. We consider this de- lows: from B. brygooi by absence of enlarged tubercles cision to be in keeping with Recommendation 24A of the around cloaca (vs. presence) and small supranasal cones Code in providing a greater universality of nomenclature. (vs. prominent); from B. decaryi and B. bonsi by absence Based on this decision, we here revise the diagnosis of of enlarged tubercles around cloaca (vs. presence); from B. antakarana, based on the original description of Rax- B. stumpffi by more dorsolateral spines (11–13 vs. 9–10); worthy & Nussbaum (1995): A Brookesia species with a from B. griveaudi by rounded supraocular cone (vs. point- complete series of 11–13, dorsolateral pointed tubercles on ed) and by more dorsolateral spines (11–13 vs. 9–10); from the body; a pelvic shield in the sacral region can be well- B. valerieae by more dorsolateral spines (11–13 vs. 9–10) defined or not well-defined, and varies in shape; no dorsal and rounded supraocular cone (vs. pointed); from B. linea ridge (keel); presence or absence of small pointed tuber- ta by absence of pointed chin tubercles (vs. four tubercles) cles on the tail, absence of prominent pointed tubercles on and typically more dorsolateral spines (11–13 vs. 11; 11 spines chin or around the cloaca; supraocular cone rounded and in B. antakarana is a rare condition). does not project further forward than nostril; the horizon- At present, the two taxa B. antakarana and B. ambreen tal distance between the snout tip and anterior margin of sis are both listed as Near Threatened in the IUCN Red List Figure 5. Haplotype networks of the mitochondrial ND2 and the nuclear Rag-1 genes in Brookesia specimens assigned morphologi- cally to B. ambreensis or B. antakarana based on pelvic shield form (see typical morphology in inset photos). Dark blue and dark red colours denote sequences referring to specimens for which detailed morphological data were available (Tables 1–2), light red and light blue colours refer to sequences of three individuals for which only field identification was available. Rag-1 haplotypes were inferred by the Phase algorithm and the network thus reconstructed with two sequences per individual. 267 Mark D. Scherz et al. (Jenkins et al. 2011a, b). We take this opportunity to briefly zahana, F. M. Ratsoavina & E. Robsomanitrandrasana revisit the conservation status of this single species follow- (2011a): Brookesia antakarana. – The IUCN Red List of Threat- ing our taxonomic revision. Recently, the borders of the ened Species, 2011: e.T172752A6911337, downloaded on 20 February 2018. Montagne d’Ambre National Park were redrawn to include a larger area and to integrate the Forêt d’Ambre Special Re- Jenkins, R. K. B., F. Andreone, A. Andriamazava, M. Anjeri- serve into the National Park (according to documents of niaina, F. Glaw, N. Rabibisoa, D. Rakotomalala, J. C. Randrianantoandro, J. Randrianiriana, H. Randriani- the Direction des Aires Protégées Terrestres of Madagas- zahana, F. M. Ratsoavina & E. Robsomanitrandrasana car, S. M. Goodman pers. comm.). Despite this improve- (2011b): Brookesia ambreensis. – The IUCN Red List of Threat- ment, the lower reaches of the park have on-going illegal ened Species, 2011: e.T172930A6943410, downloaded on 20 logging activity, and are used as a thoroughfare for local February 2018. people moving between villages. In places, this has led to Kumar, S., G. Stecher & K. Tamura (2016): MEGA7: Molecular extensive habitat degradation. These areas are largely out- Evolutionary Genetics Analysis version 7.0 for bigger datasets. side the range of B. antakarana in terms of elevation, but – Molecular Biology and Evolution, 33: 1870–1874. it can be found at low density in and around some of the Librado, P. & J. Rozas (2009): DnaSP v5: A software for compre- higher areas of disturbance (MDS, pers. obs.). We there- hensive analysis of DNA polymorphism data. – Bioinformat- fore conclude that the status of Near Threatened remains ics, 25: 1451–1452. appropriate for this species; should there be any dramatic R Core Team (2014): R: A language and environment for statisti- decline in the implementation of protection in these for- cal computing. – R Foundation for Statistical Computing, Vi- ests, the species could rapidly qualify as Endangered under enna, Austria. Available at: http://www.R-project.org/. criterion D2 (IUCN 2012). Rakotoarison, A., A. Crottini, J. Müller, M.-O. Rödel, F. Glaw & M. Vences (2015): Revision and phylogeny of nar- row-mouthed treefrogs (Cophyla) from northern Madagascar: Acknowledgements integration of molecular, osteological, and bioacoustic data re- veals three new species. – Zootaxa, 3937: 61–89. We are grateful to the following individuals: J. H. Razafindraibe, Raxworthy, C. J. & R. A. Nussbaum (1995): Systematics, spe- O. Randriamalala, R. T. Rakotonindrina, S. M. Rasolo- ciation and biogeography of the dwarf chameleons (Brookesia; njatovo, E. Z. Lattenkamp, and A. Razafimanantsoa for help- Reptilia, Squamata, Chamaeleontidae) of northern Madagas- ing us with fieldwork; the students of the 2016 BD08 (Vertebrate car. – Journal of Zoology, 235: 525–558. Morphology) course of TU Braunschweig (L. M. Bente, J. Bütt- ner, H. Fuhrmann, A. Kynzburska, S. X. Meyer, F. Nikolka, L. Raxworthy, C. J., M. R. J. Forstner & R. A. Nussbaum (2002): Raith, M. Rehbock, G. Koziel (née Reiser), M. Schrader) for Chameleon radiation by oceanic dispersal. – Nature, 415: 784– their work measuring the specimens presented here; G. Schnei- 787. der (UMMZ) for providing photographs and CT scans of the Salzburger, W., G. B. Ewing & A. Von Haeseler (2011): The holotypes of B. ambreensis and B. antakarana; D. Rabaiotti for performance of phylogenetic algorithms in estimating hap- statistical consultation; C. Hutter and one anonymous reviewer lotype genealogies with migration. – Molecular Ecology, 20: for helpful peer review and J. Köhler for additional helpful com- 1952–1963. ments; S. M. Goodman for information on the future protection Stephens, M., N. J. Smith & P. Donnelly (2001): A new sta- status of the Montagne d’Ambre region; and the Malagasy authori- tistical method for haplotype reconstruction from population ties (Direction Régionale des Forêts) for issuing the necessary re- data. – American Journal of Human Genetics, 68: 978–989. search and export permits to collect the specimens used here. Fi- Tolley, K. A., T. M. Townsend & M. Vences (2013): Large-scale nancial support was provided by the Volkswagen Foundation to phylogeny of chameleons suggests African origins and Eocene FG and MV, and by DFG (project VE 247/13-1) to MV and MDS. diversification. – Proceedings of the Royal Society of London B, 280: 20130184. References Townsend, T. M., D. R. Vieites, F. Glaw & M. Vences (2009): Testing species-level diversification hypotheses in Madagas- Bruford, M. W., O. Hanotte, J. F. Y. Brookefield & T. Burke car: the case of microendemic Brookesia leaf chameleons. – (1992): Single-locus and multilocus DNA fingerprint. – pp. Systematic Biology, 58: 641–656. 225–270 in: Hoelzel, A. R. (ed.): Molecular Genetic Analy- sis of Populations: A Practical Approach. – IRL Press, Oxford. Glaw, F. & M. Vences (2007): A field guide to the amphibians and reptiles of Madagascar. – Vences & Glaw Verlags GbR, Köln, Germany, 496 pp. ICZN (1999): International Code of Zoological Nomenclature. – The International Trust for Zoological Nomenclature, Lon- don, 306 pp. IUCN (2012): IUCN Red List Categories and Criteria: Version 3.1. – IUCN, Gland, Switzerland and Cambridge, UK. Jenkins, R. K. B., F. Andreone, A. Andriamazava, M. Anjer- iniaina, F. Glaw, N. Rabibisoa, D. Rakotomalala, J. C. Randrianantoandro, J. Randrianiriana, H. Randriani- 268